network meta-analyses in national clinical …...network meta-analyses in national clinical...
TRANSCRIPT
Network meta-analyses in national
clinical guidelines: a description and
quality assessment
Heather Stegenga, Gabriel Rogers, Prashanth
Kandaswamy, Dylan Jones
Internal clinical guidelines programme
Centre for clinical practice
National Institute for Health and Clinical Excellence (NICE)
United Kingdom
Background
• A number of treatment options
• How to decide?
• Often gaps in the literature between treatments
of interest
• Rarely one trial comparing all possible
treatments for a decision problem
• Decisions made implicitly?
Value of network meta-analysis is
making decisions more explicit that are
otherwise made implicitly
Indirect treatment comparison (ITC)
C
A B
No head-to-head trials of A vs B
Each line represents
1 or more trials
Treatment effect A vs B (indirect) = treatment effect C vs A (direct) – treatment effect C vs B (direct)
Plus, add the variances of treatment effect.
Indirect treatment comparison (ITC)
F
B
A
C
B
A
C
E
D
C
A B
Key feature: disconnected network
Mixed/multiple treatment comparison
(MTC) or multiple-treatment meta-
analysis (MTM)
C
A B
Indirect estimate of A vs B
Head-to-head trial(s) between A vs B (direct estimate)
Pool indirect and direct estimates to get more precise estimate of treatment
effect
Mixed/multiple treatment comparison
(MTC) or multiple-treatment meta-
analysis (MTM)
Important!
Need to check difference (or
consistency) between indirect and
direct estimates
Inconsistency = imbalance in effect
modifiers in indirect and direct
studies
Why combine direct and indirect?
• Helpful if there is some reason that we might not be confident in the results of the head-to-head study/studies: – Results are not precise
– Low study quality
– Manufacturer supported
– Study sizes
– Small number of trials (ie. only 1 head-to-head trial)
• Level of consistency/inconsistency across the network (between direct and indirect treatment effects) can be informative and provide coherence throughout the whole network of treatments
A developing methodology = different
terminology
Indirect
treatment
comparison
(ITC)
Network
meta-
analysis
Multiple
treatment meta-
analysis (MTM)
Multiple
treatment
comparison
(MTC)
Mixed
treatment
comparison
(MTC)
A developing methodology = different
terminology
Indirect
treatment
comparison
(ITC)
Network
meta-
analysis
Multiple
treatment meta-
analysis (MTM)
Multiple
treatment
comparison
(MTC)
For the purposes of this project,
mixed/multiple treatment
comparisons AND indirect
treatment comparisons are both
considered network meta-analyses
(NMA)
Mixed
treatment
comparison
(MTC)
NICE Decision support unit checklist –
4 main areas
A. Definition of decision problem
B. Method of analysis
C. Issues specific to the network synthesis
D. Embedding the synthesis into a probabilistic cost-
effectiveness model
Our objective
• To describe and quality assess the NMAs used in NICE
clinical guidelines
• Adapted version of checklist – added some descriptive
variables to provide some points of comparison
Inclusion criteria: published up to January 2012
Exclusion criteria: in development (including those in
the public domain which have been out for public
consultation)
NICE clinical guidelines
Total 145 published guidelines
8% (11) with NMA
-7 de novo
-3 from published study
-1 with both
Number of networks per guideline
5 3
1
15
1 3
2
6
2 1
3
02468
10121416
Type of interventions
55% 27%
18% pharmacological
various (includingpharmcological)
endoscopic (includingradiotherapy, etc) orsurgical procedures
Use in economic model
0
1
2
3
4
5
6
7
8
9
10
Used in economic model
unclear
Methodology
0% 10% 20% 30% 40% 50% 60% 70% 80%
Bayesian
Bucher's adjusted indirect comparison
"Adjusted indirect comparison"
Comparison of point estimates
unknown
Re: methods
Complexity of networks
Average (range)
Interventions in decision
space
8.4 (2-19)
Additional interventions in
network (for indirect
comparisons)
2.2 (1-5)
Number of connected loops
(MTC)
6.2 (1-14)
Interventions with only one
link
29% (0-55%)
Number of trials 41 (2-128)
Trials in head-to-head
comparison (MTC) (vs
indirect)
78% (50-100%)
Links with only one trial 47% (0-100)
Presentation of NMA
• Diagram/picture of the network:
– All but 1 guidelines with de novo NMA
(1 without was an ITC)
Schizophrenia (CG82)
Re: presentation
– 3 of the 8 guidelines with de
novo networks, including the
2 most recent
– Use with caution: doesn’t take
into account uncertainty
around treatment effect
Colorectal cancer (CG131)
Re: presentation
– Only one guideline used a rankogram
– Rankograms which take into account the probability
of being best, second best, etc are ideal
Depression (update) (CG90)
Inconsistency (MTC)
• Recommended methods:
– Repeat indirect methods for each ‘loop’
– Develop a model of inconsistency and compare this
with the standard network inconsistency
• Some guidelines said they checked for inconsistency
but it was often not clear how
• Of the 122 possible inconsistencies across the
networks, there were only 2 reported inconsistencies (2
different guidelines).
• There was very little discussion about inconsistencies in
many of the guidelines. Does that mean that few were
found?
Overall difficulties encountered
• Considerable variation of where details found
• Some variation expected ie. different developers,
different topic areas and, importantly, developing
methodology
• Methods were not reported or were unclear
• Checklist is new (available in early 2012) and is an
attempt to improve methods and reporting
Discussion
• Value of network meta-analysis is being explicit about
decisions otherwise made implicitly
• Value of being explicit is lost when description is lost
• Inability to assess the adequacy of decisions made from
the results of the NMA
Recommendations:
Fully describe methods used
Present networks in diagram
Use rankograms
Report and discuss inconsistency
across networks
Further information
• All guidelines available from www.nice.org.uk
• NICE Decision support unit: http://www.nicedsu.org.uk/
My contact details:
Disclosure of Interests (last 3 years)
Heather Stegenga
I certify that, to the best of my knowledge, no aspect of my current personal or
professional situation might reasonably be expected to affect significantly my
views on the subject on which I am presenting.